Home > News content

Technological companies like Apple, Google and other popular neuroscience companies are rushing in

via:博客园     time:2019/6/20 15:37:44     readed:149


Boston's Tim

Tencent Science and Technology News, June 20, according to foreign media reports, with Apple, Facebook and Amazon and other technology companies competing to develop artificial intelligence and brain-computer interface technology, neuroscientists on brain activity related research has attracted more and more attention. Because of better pay and better opportunities, many neuroscientists choose to jump from academia to science and technology to engage in basic research, using their deep understanding of brain activity as technology application services.

Jaguar is a mouse. It lives at Harvard's Roland Institute (Harvard).

To track Jaguar's brain activity, the researchers genetically modified it so that neurons in its brain emit fluorescence when they discharge. The microscope at the top of the platform records the luminous images of the brain as Jaguar plays.


Neuroscientists at Harvard's Roland Institute studied mice and trained them to perform tasks including video games and mobile joysticks.

Over the past few decades, Marcis's research has only advanced human understanding of rat and brain function. Today, however, more and more professional animal researchers are helping to develop artificial intelligence software and brain-computer interface, and Marcis is just one of them. She wants to find out how mice learn, partly because it tells humans how to teach computers to learn. For example, watching mice respond to unexpected situations in video games may one day enable humans to teach similar skills to robots.

Other neuroscientists are studying the singing skills of zebra sparrows. And some people are becoming experts in the electrical conductivity of sheep skulls. More scientists choose to study fruit flies or worms, which have relatively simple neural structures. Over the past few years, large technology companies have been poaching talent from universities. Apple, Facebook, Google and Twitter all hired Ph.D. students from a scholarship program they led, Mr. Mathis said:

Of course, animals have long played an important role in promoting scientific research in enterprises, especially in the medical field. But the technological leap is entirely different if you want to convert the anatomy of the zebra finch's voice organ into Siri speech recognition software, or if you want to turn mouse games into Amazon's Quan'an Android software. As the challenges of the whole new industry continue, the competition to uncover the secrets of animal thinking becomes more and more peculiar.

1958, Cornell University neurobiologist Frank

Even today, this generalization exaggerates the overlap between computational and cognitive domains. It's quite difficult for scientists to copy things that they don't really understand. The working principles of the brain, such as how neurons store memories, remain a mystery to neuroscience, so the digital counterparts of human brain neurons can only be defective imitations. Their trained primary processing engines can perform a great deal of statistical computation and pattern recognition. Although they get a biological name, they are still far from the real meaning.


Rats are helping researchers uncover the secrets of neural networks

Nevertheless, as the technology industry pursues so-called General Artificial Intelligence (AGI), the barriers between the two fields become less unshakable. The goal of AI is a perceptual machine that can solve problems on its own, rather than relying on human beings to train it. To the comfort of some ethicists, we still have a long way to go from AI in general, but many computer scientists and neuroscientists bet that brain research will show us the way.

In addition, several companies are trying to build brain-computer interfaces, such as helping artificial limbs move like natural limbs, or allowing people to download knowledge into their brains. Silicon Valley star entrepreneur Elon

Marcy's lab is conducting basic research on these technologies.

Corresponding studies, including virtual box games and a game that looks like a Mario racing car, seem to be more difficult. For the latter, a mouse sat straddling two custom electric discs with its paws embedded in grooves on both sides. The screen shows a green channel with a blue rectangle at the end. The experimental rat was driving close to the blue rectangle, and it had to drive carefully to keep on the virtual path. Like humans, these mice have dull eyes when they play. The whole experiment lasts about half an hour.

The microscope will look at the mouse's brain and record the incredible amount of information. Marcy said:

One of the main purposes of Masses'research is to learn more about how animals adapt quickly to changes in the physical environment. For example, when you pick up an object of unknown weight, your brain and body quickly calculate how much force it needs to exert. At present, robots are unable to do this, but robots injected with mouse neuron learning mode are likely to do so. Marcis says mice can help bridge the gap. Their brains are complex enough to make high-level decisions, but they are also simple enough for researchers to infer these connections in a reasonable time.

It was not until recently that powerful computers were developed to collect, process and analyze the amount of data generated by a small portion of about 75 million neurons in the mouse brain. Artificial intelligence software has made great progress in recent years, thus automating most of the research work. Mathis and her husband Alex.


Mackenzie and Alex.

Scientists used to rely on manpower to do similar work. Now, it takes only a few minutes for the software to complete tasks that took weeks or months of manual work. Alex said:

Such software development and data analysis have also attracted the interest of technology companies in neuroscientists, just as they need to understand animal cognition. Modern brain researchers must know how to code and process a lot of information, just as Google's artificial intelligence workers improve advertising algorithms or allow autopilot cars to connect. Animal-centered neuroscientists are also used to using unconventional research methods.

[人名] Teem

In the late 1990s, Ochi majored in mechanical engineering at (Georgia Institute of Technology), Georgia Institute of Technology, and worked for a company specializing in automation plant systems research and development. His job is to teach robots to identify gadgets or auto parts and categorize them on conveyor belts. He said:

A songbird like the zebra finch has an unusual skill. Although most animals instinctively know how to make sounds, zebra finches imitate the sounds they hear and then change their tunes, seeming to have some semantic understanding of the tunes they hear. Decades of research have identified the structure of the brain responsible for this behavior in finches, known as the vocal nucleus. Researches in this field have enabled scientists to have a deeper understanding of the functions of neural circuits, and also provided information for other studies of human motion, sensation and emotion. Understanding how birds mimic each other's calls can help explain how to do the same thing, which is important for training machines to master language skills.


Birds have a semantically similar understanding of song, and if scientists can understand it correctly, they can be applied to speech recognition software.

At a birdhouse at Boston University, Ochi works with about 300 birds. In one experiment, the researchers will equip a zebra sparrow with a miniature backpack with batteries that can power a large number of electronic devices attached to the skull. The bird was then placed in a microwave oven-sized soundproof box that sounded for days in a row, while Ochi and his team looked at the brains of zebras through mechanisms similar to those used by Masis in mice. As researchers learn more about the vocal nucleus of the sparrow, the brain of the lark becomes more and more accurate.

Before Ozzie took charge of Gardner Labs, his predecessor, Tim,


Ouch is in the Boston University lab.

Songbird researchers are one of the hottest employees in the field of artificial intelligence. After working for a while at the University of California, Berkeley, Channing? Chning Moore has joined Google's voice-understanding team to develop a sound recognition system that is as complex as Google's image recognition software, which can distinguish the sound of the alarm and the baby's cry. In Intel, another doctor at the University of Berkeley, Taylor,

Fredrik, a professor at the University of Berkley

Since the 1970s, scholars have been trying to declare that this is the age of neuroscience, but at the beginning of this century, the prospects of young neuroscience graduates are dim, and the number of further studies in related majors is low. According to the U.S. Department of Education, 15 years ago, fewer than 1,500 undergraduate neuroscience graduates and fewer than 400 doctorates were awarded. Even in this case, schools are unable to provide them with sufficient full-time jobs or grants.

When Drew Robson graduated from Princeton University and earned a degree in mathematics in 2005, his undergraduate teacher gave him a memorable suggestion:Don't do neuroscience whatever you do. Robson ignored it, but with his partner, lover Jennifer.


[人名] Drue

Rawley's team studied zebrafish, members of the minnow family, whose bodies were transparent when they were young, allowing researchers to directly observe their neurons. Robson and Lee invented a special mobile microscope to help them record which neurons are active when fish swim. To capture different aspects of zebrafish behavior, researchers alter neuronal electrical stimulation, causing zebrafish to turn or continue to accelerate forward swimming.

Like many of their peers, Robson and Lee know a lot about the relationship between brain science and artificial intelligence technology. Last year, the couple bought a Tesla electric vehicle, and their professional attainments were pleased to see the continuous development of the autopilot system. When Tesla avoids other vehicles, it recalls zebrafish strategies, such as switching from hunting mode to fast swimming mode when they find predators. As Tesla tries to upgrade autopilot technology from basic object recognition to human-like decision-making, Robinson and Lee's in-depth understanding of such behavior may someday improve Tesla's neural network.


The body of young zebrafish is transparent, which provides a way to map complex brain maps and reproduce the process by machine.

In the field of neuroscience, the soft-border between public and private companies raises a question: who will ultimately control the possible merger between humans and the machine. University of advanced research has long been competing with technology companies with more powerful computers and more data sets. A person who has just obtained a doctor's degree is only about $50,000 a year at an ordinary university, while the private enterprise offers more than 6-digit annual salary, or even higher. Chris

In addition to salaries, many neuroscientists are attracted to the private sector because it often gives them an opportunity to do more exciting or even strange jobs without writing grant applications. However, switching to Silicon Valley may also mean disrupting promising research routes or letting colleagues drift with the tide. When Gardner left the lab to work in Neuralink, one of his PhD students had to transfer.

Lee and Robson will travel to Max in Tubingen, Germany, in September this year.


Robson and Lee are next to the NMR imager.

Robson and Lee say the development of artificial intelligence and brain-computer interface technology will force humans to become more humanized. After all, if one of our goals is to inculcate our own morality in machines, we will have to focus more on what morality is than ever before. Who should have the ability to think more? Should self-driving cars choose to save passengers or pedestrians? How smart should the machine be?

China IT News APP

Download China IT News APP

Please rate this news

The average score will be displayed after you score.

Post comment

Do not see clearly? Click for a new code.

User comments